On Tue, 7 Jan 2003, Martin Gasbichler wrote:
> >>>>> "RT" == RT Happe <rthappe@web.de> writes:
> RT> The important issue is that the buffer size doesn't determine whether the
> RT> port can satisfy the request from the buffer. The port-index or
> RT> fill-pointer of the buffer does.
>
> The code wasn't there to test whether there is something in
> the buffer or not but to guess whether the port's policy is
> bufpol/none.
Yes, I remember a source comment to that effect. But whatever the buffer
policy, read-block 'immediate won't (always) do since it may fail to
deliver any byte. (I didn't look up the port-handler, but I assume it
gives up on a locking device if asked for 'immediate bytes. This would
fit the description in ports.scm or somewhere.) That is, r.s.p wouldn't
implement the conditional forward-progress guarantee of the spec.
rthappe
|